2. ENDOLEAK DETECTION AND CLASSIFICATION AI

CHALLENGE! Create an AI model that analyzes post-EVAR CT scans to detect and classify endoleaks (Type I, II, III, IV). The system shall be able to:

• identify and segment endoleak areas visible on CT, determining the exact location (around the graft, in the aneurysm sac, in the surrounding vessels);
• automatically classify the endoleak type (I, II, III, IV) based on its morphology and location;
• provide a detailed report showing the leak size and severity in relation to the surrounding vascular geometry, useful for follow-up and monitoring.

EXPECTED IMPACT: a type II endoleak occurs when, after the implant of a prostheses, small peripheral arteries keep on supply blood to the aneurysmal sac, which becomes unstable and threaten the health of an operated patient. This problem has been addressed by preventive actions during the implant of the prosthesis, such as embolization of the sac with thrombogenic material (but this is not time-efficient) or by filling the sac with a polymer (but this proven to fail due to the mass of the material and mismatch of mechanical properties with the sac). Prostheses with peptide filaments functionalized with amino acids (with a great ability to coagulate blood) have been explored, but not industrialized.

3. CFD SIMULATION FOR BLOOD FLOW AND WALL STRESS PREDICTION

CHALLENGE! Develop an artificial intelligence algorithm that, using pre-operative CT scans, integrates patient-specific vascular geometry with CFD-derived hemodynamic biomarkers to predict the risk of post-EVAR adverse events. The model will combine anatomical features and flow-related parameters – such as Wall Shear Stress (WSS), Oscillatory Shear Index (OSI), and Relative Residence Time (RRT) – to generate a patient-specific risk classification, supported by interpretable maps highlighting anatomical regions and flow profiles potentially associated with complications such as endoleaks, graft migration, rupture, or reintervention.

EXPECTED IMPACT:  A pre-operative CT-based risk stratification tool with direct applications in EVAR planning and clinical management:

• improving patient selection;
• optimizing procedural strategy;
• personalizing post-operative surveillance and follow-up.

4. LLM-BASED SUMMARIZATION OF VASCULAR SURGEONS’ POST-EVAR CLINICAL REPORTS 

CHALLENGE! Develop a LLM model that, starting from the textual report written by vascular surgeons during the first post-EVAR clinical visit, generates an automatic summary of key information. The system must identify and summarize clinically relevant issues such as suspected endoleak, signs of prosthesis migration, unintentional coverage of arterial branches or other early complications. The goal is to support structured follow-up and interdisciplinary communication, simplifying the consultation of textual reports.

EXPECTED IMPACT: the usefulness could be in the follow-up visits after the first, in order to quickly and clearly have the key information on any complications so as not to lose sight of them due to carelessness, also optimizing the time for the surgeon and boosting interdisciplinary communication.

5. COMPREHENSIVE SYNTHETIC CT IMAGING LIBRARY THROUGH GENERATIVE AI